Exploring the optimal range of blood pressure in elderly sepsis patients: a retrospective cohort study based on MIMIC-IV data Authors: Yuanwen Ye, Feifei Li, Liangen Lin, Linglong Chen, Yuequn Xie, Wang Lv, Baohua Yang Published: 2025 ABSTRACT AbstractObjectiveTo explore the optimal range of blood pressure (BP) in elderly sepsis patients.MethodsA retrospective cohort study was conducted using data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV). Restricted cubic spline (RCS) analysis was employed to examine and visualize the nonlinear relationship between BP and the incidences of in-hospital mortality and atrial fibrillation. Optimal systolic blood pressure (SBP), diastolic blood pressure (DBP), and mean arterial pressure (MAP) ranges were identified, and their association with 28-day mortality was validated using Cox regression analysis, propensity score matching (PSM), inverse probability weighting (IPTW), doubly robust model estimation (DR), and Kaplan–Meier survival curves (K-M).ResultsA total of 2,253 patients were included. RCS analysis revealed an L-shaped relationship between BP and in-hospital mortality, while the risk of atrial fibrillation increased above certain upper limits. The optimal SBP, DBP and MAP ranges were 108–118, 51–57, and 69–74 mmHg, respectively. Further statistical models confirmed that patients within the optimal BP range were associated with lower 28-day mortality. K–M survival analysis demonstrated that patients within the optimal BP range had a higher probability of survival (All Log-Rank P < 0.05).ConclusionsThis single-center study suggests an optimal BP range for elderly sepsis patients that is associated with reduced 28-day mortality. These findings point to a potential optimal range for hemodynamic management in this population and warrant validation in prospective, multi-center studies.Clinical trialNot applicable.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12879-025-12255-w.Keywords: Sepsis, Elderly, Blood pressure, 28-day mortality, Optimal range Introduction Sepsis poses a significant threat to life, marked by elevated morbidity and mortality rates, as well as substantial treatment costs, thereby exerting considerable pressure on healthcare systems worldwide and adversely affecting patient well-being [1]. The optimization of sepsis management, particularly within critical care environments, remains a formidable challenge [2]. Blood pressure (BP) is a crucial physiological parameter in the management of sepsis, as maintaining it within appropriate ranges ensures adequate organ perfusion and improves patient outcomes [3]. According to the 2021 guidelines from the Surviving Sepsis Campaign, adult patients experiencing septic shock who are on vasopressors should target an initial MAP of 65 mmHg [4]. Nonetheless, individualized BP management may be warranted for different patient populations. For example, research by Martin Dünser et al. suggests that a MAP exceeding 75 mmHg may be necessary to achieve adequate renal perfusion in septic patients [5], while Lina Zhao et al. advocate for higher BP targets in hypertensive patients compared to normotensive individuals, recommending an optimal MAP of 70–80 mmHg (65–73 mmHg for those without hypertension) and DBP of 54–62 mmHg (50–60 mmHg for non-hypertensive patients) [6]. Despite the well-documented association between BP control and sepsis prognosis, the majority of research has concentrated on general septic populations, with limited investigation into optimal BP targets for elderly patients [7]. Given the global increase in the aging population, the incidence of sepsis among the elderly is rising, accounting for up to 60% of cases in certain studies [8]. These patients often exhibit reduced physiological reserves, multiple comorbidities, and impaired vascular regulation, rendering them more vulnerable to BP fluctuations during septic shock. Failure to address these specific needs may elevate the risk of organ failure, a significant contributor to the high mortality rate in this demographic [9–11]. Consequently, the development of individualized BP management strategies tailored to the unique physiological and pathological characteristics of elderly patients is imperative. This retrospective cohort study utilizes the MIMIC-IV database to examine clinical outcomes in elderly septic patients across varying BP levels, with the objective of identifying an optimal BP “golden zone” that enhances prognosis. Through rigorous data analysis, the study aims to establish evidence-based BP targets for elderly patients undergoing hemodynamic support, thereby providing a theoretical basis for personalized treatment strategies. By addressing this research gap, the findings are expected to offer significant insights for future prospective studies and clinical practice, ultimately contributing to the reduction of morbidity and mortality in elderly septic patients and advancing global public health initiatives. Materials and methods Study settings This retrospective cohort study was based exclusively on data obtained from the MIMIC-IV database (MIMIC-IV 2.0 version). Researchers completed the “Protecting Human Research Participants” course and obtained ethical approval from BIDMC and the Massachusetts Institute of Technology (MIT) (Record ID: 43546933). Patients Patients meeting the Sepsis-3.0 diagnostic criteria were enrolled in the study. The exclusion criteria encompassed repeated admissions, an Intensive Care Unit stay less than 48 h (to ensure adequate data collection for hemodynamic assessment and to reduce the impact of very early death or discharge), admission age less than 75 years, lack of vasopressor therapy drugs during the ICU stay, and a documented history of atrial fibrillation. Data collection Data extraction was performed using PostgreSQL. Variables extracted included demographic information, coexisting illnesses, vital signs, laboratory parameters, critical illness scoring systems, and clinical treatment information. Demographic information Demographic information included age and gender. Coexisting illnesses Coexisting illnesses encompassed myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatic disease, peptic ulcer disease, mild liver disease, diabetes without complications, diabetes with complications, paraplegia, renal disease, malignant cancer, severe liver disease, metastatic solid tumor, and acquired immune deficiency syndrome. Additionally, the Charlson Comorbidity Index (CCI) was calculated to assess the impact of comorbidities. Vital signs Vital signs included temperature (T), systolic blood pressure (SBP_ Non Invasive), diastolic blood pressure (DBP_ Non Invasive), mean blood pressure (MAP_ Non Invasive), heart rate (HR), respiratory rate (RR), and SpO₂. For analysis, the median values of vital signs measured during hospitalization were used as the measure of central tendency for each patient. Laboratory parameters Laboratory parameters included white blood cell count (WBC), hemoglobin (Hb), platelet count (PLT), creatinine, blood urea nitrogen (BUN), potassium, sodium, chloride, calcium, glucose, bicarbonate, and lactate. For analysis, the median values of laboratory parameters during hospitalization were used as the measure of central tendency for each patient. Severity scoring systems Severity scoring systems included the Glasgow Coma Scale (GCS) and Sequential Organ Failure Assessment (SOFA). For analysis, the median values of these scores during hospitalization were used as the measure of central tendency for each patient. Clinical treatment information Clinical treatment information included the use of renal replacement therapy (RRT), mechanical ventilation (MV), occurrence of acute kidney injury (KDIGO Stage 3), length of hospital stay, length of ICU stay, atrial fibrillation, in-hospital mortality, and 28-day mortality. The primary study variables were SBP, DBP and MAP during hospitalization, with the median values used to represent each patient’s BP levels during hospitalization. When using restricted cubic spline (RCS) analysis, the primary clinical events were atrial fibrillation and in-hospital mortality; when using statistical modeling, the clinical event was 28-day mortality. Atrial fibrillation was defined as any history of atrial fibrillation occurring during hospitalization. In-hospital mortality was defined based on survival status at discharge, and 28-day mortality was defined as the survival status 28 days after admission. To mitigate potential bias due to missing data, variables with more than 30% missing values were excluded from subsequent analyses. For variables with less than 30% missing values, multiple imputation (MI) methods were employed. Statistical analysis The Shapiro–Wilk test was used to assess the distribution of the data. Continuous variables with non-normal distributions were analyzed and are presented as the median and interquartile range (IQR), while categorical variables are expressed as counts and percentages. Continuous variables between the two groups were compared using nonparametric tests, and categorical variables were compared using Fisher’s exact test. Restricted cubic spline analysis was employed to examine and visualize the nonlinear relationship between BP levels during hospitalization and the incidences of in-hospital mortality and atrial fibrillation. Based on these results, optimal ranges for SBP, DBP and MAP were identified, and the study cohort was divided into two subgroups with different BP levels for subsequent analyses. Various statistical models were employed to confirm the association between BP levels and 28-day mortality. Lasso regression + Cox regression analysis To reduce multicollinearity among variables, a Lasso regression model was used to select covariates. Subsequently, the relationship between optimal BP range and 28-day mortality was tested using a Cox regression model. Propensity Score Matching Propensity scores were calculated using a gradient boosted model (GBM), which minimized differences in covariates between the two groups. A matched cohort was then obtained through 1:1 matching based on these propensity scores, and Cox regression analysis was performed on the matched cohort. Propensity score IPTW An inverse probability weighting (IPTW) model was used to generate a weighted cohort using estimated propensity scores as weights, and Cox regression analysis was performed on the weighted cohort. Doubly robust model Estimation The relationship between optimal BP range and 28-day mortality was further explored in the matched cohort using a doubly robust analysis that combined multifactorial regression with propensity score adjustment. The Kaplan–Meier (KM) method was employed to plot survival curves, and the log-rank test was used to compare differences between groups. A series of sensitivity analyses were conducted to further validate the robustness of the study results. All statistical analyses were performed using R 3.5.0 software, with a significance level set at P < 0.05. Results Patient selection A total of 32,971 patients diagnosed with sepsis were identified from the MIMIC-IV database. Following the application of exclusion criteria, 30,718 patients were omitted, resulting in a final study cohort comprising 2,253 patients. This cohort was stratified into a survival group (n = 1,737) and a mortality group (n = 516) based on the occurrence of in-hospital mortality events. Furthermore, the cohort was categorized into an atrial fibrillation group (n = 1,087) and a non-atrial fibrillation group (n = 1,166) according to the presence or absence of in-hospital atrial fibrillation. The patient screening process is depicted in Fig. 1. Flow chart for patient selection Baseline characteristics Appendix Table 1 provides a detailed account of the baseline characteristics of the patients (Appendix Table 1 Baseline characteristics of patients with sepsis). Compared to the sepsis-survival group, patients in the mortality group were older, had a higher prevalence of comorbidities, exhibited increased heart rates and respiratory rates, elevated pulse oximetry readings, higher Sequential Organ Failure Assessment (SOFA) scores, and greater utilization of renal replacement therapy (RRT) and mechanical ventilation (MV). Additionally, they experienced a higher incidence of acute kidney injury (AKI) at KDIGO Stage 3 during hospitalization. Notably, the mortality group had shorter durations of hospital and intensive care unit (ICU) stays.Laboratory analyses demonstrated significant differences between the two cohorts, with the exception of blood sodium levels. In comparison to the sepsis-survival group, patients in the mortality group exhibited lower BP during hospitalization (SBP: 116 [109–124] vs. 110 [104–119] mmHg, p < 0.001; DBP: 54.5 [50.0–59.0] vs. 53.0 [48.0–58.0] mmHg, p < 0.001; MAP: 72.0 [68.0–77.0] vs. 70.0 [66.0–75.0] mmHg, p < 0.001). In a comparative analysis between patients with atrial fibrillation (AF) and those without, individuals in the AF cohort demonstrated a significantly elevated utilization of renal replacement therapy (RRT) and mechanical ventilation (MV), a higher incidence of acute kidney injury (AKI) at KDIGO Stage 3, prolonged durations of hospital and intensive care unit (ICU) stays, and an increased rate of in-hospital mortality. Although the difference in SBP was not statistically significant (SBP: 114 [107–123] vs. 115 [107–124], p = 0.196), the AF group exhibited significantly higher DBP (DBP: 53.0 [49.0–58.0] vs. 55.0 [51.0–60.0], p < 0.001) and MAP (MAP: 71.0 [67.0–75.0] vs. 73.0 [68.0–77.5], p < 0.001) compared to the non-AF group. Restricted cubic splines analysis to estimate the BP targets for elderly sepsis patients In this study, we investigated the nonlinear associations between SBP, DBP, MAP, and two clinical events: in-hospital mortality and atrial fibrillation, utilizing restricted cubic spline (RCS) analysis. The findings revealed an L-shaped relationship between SBP, DBP, MAP, and in-hospital mortality, suggesting that lower BP levels are correlated with increased in-hospital mortality among elderly patients with sepsis. Mortality rates decreased progressively with rising BP. Specifically, in-hospital mortality was significantly reduced when SBP was ≥ 108 mmHg (see Fig. 2A), DBP was ≥ 51 mmHg (see Fig. 2B), and MAP was ≥ 69 mmHg (see Fig. 2C). Restricted Cubic Splines analysis to estimate the relationship between SBP, DBP, MAP and incidence of In-hospital mortality and incidence of atrial fibrillation in elderly sepsis patients. A. The relationship between SBP and incidence of In-hospital mortality and incidence of atrial fibrillation. B. The relationship between DBP and incidence of In-hospital mortality and incidence of atrial fibrillation. C. The relationship between MAP and incidence of In-hospital mortality and incidence of atrial fibrillation Conversely, when analyzing the association between BP and atrial fibrillation, we observed that the incidence of atrial fibrillation markedly increased when SBP exceeded 118 mmHg (see Fig. 2A), DBP exceeded 57 mmHg (see Fig. 2B), and MAP exceeded 74 mmHg (see Fig. 2C). Based on the RCS analysis, the optimal BP ranges for elderly sepsis patients were determined to be: SBP 108–118 mmHg, DBP 51–57 mmHg, and MAP 69–74 mmHg. Prognostic analysis of BP targets and elderly sepsis patients To further evaluate the impact of target values for SBP, DBP and MAP on patient prognosis, specifically 28-day mortality, as informed by the results of the restricted cubic spline (RCS) analysis, we utilized four distinct statistical methodologies: conventional statistical analysis, propensity score matching (PSM), inverse probability weighting (IPTW), and doubly robust modeling. Consistently, across all four statistical approaches, SBP (108–118 mmHg), DBP (51–57 mmHg), and MAP (69–74 mmHg) were associated with a lower risk of 28-day mortality in elderly patients with sepsis (see Figure 3). Application of multiple statistical models confirmed the relationship between optimal range of BP and 28-day mortality of elderly sepsis patients In the conventional statistical analysis model, to mitigate multicollinearity among variables, we employed the Lasso regression model to screen all variables, excluding SBP, DBP and MAP, to identify potential risk factors influencing 28-day mortality. The process of variable selection is depicted in Appendix Figure 1 A. Appendix Fig. 1B illustrates that, following the exclusion of certain covariates, two models were derived: the dashed line on the left represents the minimal model, which includes 37 variables, while the dashed line on the right represents the streamlined model, comprising 14 variables. Subsequently, SBP, DBP and MAP were incorporated as independent variables into the Cox regression model, with 28-day mortality serving as the dependent variable and the 14 variables identified through Lasso regression as covariates. In the propensity score matching model, we matched 38 covariates (excluding SBP, DBP and MAP) in a 1:1 ratio to address or reduce imbalances between the BP target group (SBP 108–118 mmHg, DBP 51–57 mmHg, and MAP 69–74 mmHg) and the non-target group. Appendix Fig. 2 demonstrates the effective balance achieved between groups post-propensity score matching. To further enhance group balance, we combined propensity score matching with inverse probability of treatment weighting (IPTW) to adjust for intergroup differences. The balance of variables following IPTW is depicted in Fig. 4. The standardized mean difference (SMD) of the original cohort was compared with the IPTW cohorts, which showed that covariates were well balanced between classes after IPTW (< 0.1). SMD stands for standardized mean difference, and IPTW refers to inverse probability of treatment weighting.(A). SBP; (B). DBP; (C). MAP Kaplan–Meier survival curves were employed to illustrate the prognostic disparities between elderly sepsis patients within the BP target group ( [SBP] 108–118 mmHg, [DBP] 51–57 mmHg, [MAP] 69–74 mmHg) and those outside this target group, as depicted in Fig. 5. Furthermore, subgroup analyses indicated that maintaining BP within the specified target range (SBP 108–118 mmHg, DBP 51–57 mmHg, and MAP 69–74 mmHg) served as a protective factor against 28-day mortality in elderly sepsis patients across most subgroups, as presented in Appendix Table 2 (Appendix Table 2 The results of the subgroup analysis of the optimal range). The Kaplan–Meier survival curves of the BP target group and the no BP target group, based on Restricted Cubic Splines analysis. A. The survival curves of SBP. B. The survival curves of DBP. C. The survival curves of MAP Discussion The findings of this study indicate that the optimal BP ranges for elderly patients with sepsis are 108–118 mmHg for SBP, 51–57 mmHg for DBP, and 69–74 mmHg for MAP. Multiple statistical models have corroborated the association between maintaining BP within these specified ranges, revealing a significant association with decreased 28-day mortality rates. These results imply that this range may be associated with optimal outcomes for elderly sepsis patients. BP management is a fundamental aspect of sepsis treatmen [12–15]. The 2021 Surviving Sepsis Campaign (SSC) guidelines recommend an initial MAP target of 65 mmHg for adult patients with septic shock receiving vasopressors [4]. However, emerging evidence, along with our findings, indicates that a standardized approach may be insufficient, particularly for the elderly population. Our study identified a marginally higher optimal MAP threshold of 69–74 mmHg for elderly patients with sepsis. The requirement for elevated MAP targets in this demographic may be attributed to various age-related physiological and pathophysiological mechanisms. Firstly, vascular structural changes, such as atherosclerosis, often result in reduced physiological reserve and impaired vascular compliance in elderly patients [9–11]. This diminished autoregulatory capacity compromises their ability to maintain adequate organ perfusion at lower BP levels, rendering them vulnerable to perfusion deficits in organs such as the kidneys and brain, even at MAP levels exceeding the conventional 65 mmHg threshold [16, 17]. Secondly, age-related changes in sympathetic nervous system activity and baroreceptor sensitivity may attenuate the body’s hemodynamic response to sepsis and vasopressors, thereby necessitating higher perfusion pressures to ensure adequate blood flow to vital organs [18, 19]. Furthermore, chronic conditions commonly observed in the elderly population—often pre-existing yet potentially subclinical, such as hypertensive nephrosclerosis and cerebrovascular disease—may require an increased MAP to surpass elevated critical occlusion pressures within these vascular territories, thereby mitigating the risk of occult ischemic injury [20, 16]. Consequently, although a MAP of 65 mmHg may suffice for younger and healthier individuals, our findings indicate that elderly patients might benefit from a moderately elevated MAP range. This adjustment could ensure adequate microcirculatory perfusion and oxygen delivery, thereby diminishing the risk of progressive organ dysfunction and enhancing survival rates. Our analysis, conversely, identified an association between higher BP levels and an increased incidence of atrial fibrillation, highlighting the necessity for precise hemodynamic management. While hypotension is detrimental, excessively high BP—potentially resulting from high-dose vasopressors or fluid overload—elevates ventricular wall stress and myocardial oxygen consumption, which may trigger atrial fibrillation in susceptible individuals [21, 22]. This issue is particularly critical in elderly patients, who often present with diastolic dysfunction or subclinical cardiac disease.The optimal BP range identified in this study (SBP 108–118 mmHg, DBP 51–57 mmHg, MAP 69–74 mmHg) therefore represents a range that is associated with both lower mortality and a lower incidence of atrial fibrillation. This range appears to balance the risks associated with both lower and higher BP levels. Our findings align with and expand upon existing research. While the Survive Sepsis Campaign guidelines offer universal targets, recent studies increasingly highlight the necessity for individualized BP management [16, 17, 21]. Lamontagne et al. observed that although the standardized MAP target of 65 mmHg is broadly applicable, its appropriateness for elderly patients remains a subject of debate [16]. Our findings provide evidence supporting the need for moderately elevated BP targets in this vulnerable demographic. Additionally, by demonstrating an association between elevated BP levels and atrial fibrillation—which may reflect the interplay between hemodynamic stress and the proinflammatory, proarrhythmic state of sepsis—our data enhance the understanding of sepsis-associated atrial fibrillation. However, it is also important to consider that atrial fibrillation itself can cause hemodynamic instability and BP fluctuations, or that both high BP and AF may be driven by shared underlying factors such as heightened sympathetic tone or systemic inflammatio [22–24]. Therefore, the observed association should be interpreted as correlational rather than causal. This study presents several significant strengths and contributions to the existing body of research on BP management in elderly patients with sepsis. Firstly, the study utilizes the high-quality MIMIC-IV database to ensure robust real-world data. Secondly, it employs advanced statistical methodologies, including restricted cubic splines (RCS), propensity score matching (PSM), inverse probability of treatment weighting (IPTW), and doubly robust estimation (DR), to enhance the robustness of the findings and mitigate confounding biases. Thirdly, the study proposes an optimal mean arterial pressure (MAP) range of 69–74 mmHg, which is higher than the standard 65 mmHg, thereby challenging existing guidelines and promoting personalized treatment approaches. Despite its contributions, the study is not without limitations. As a retrospective cohort study utilizing the MIMIC-IV database, it is susceptible to selection bias and residual confounding. Specifically, the exclusion of patients with an ICU stay of less than 48 h, while necessary to ensure sufficient data for hemodynamic profiling and to mitigate the impact of rapid mortality or transfer, may have introduced selection bias by excluding those with the most fulminant or rapidly resolving courses of illness. Although we employed comprehensive statistical adjustments, including propensity score methods, to balance observed covariates between groups, unmeasured variables—such as fluid balance, precise medication dosages, and specific etiologies of sepsis—may still affect the findings. Furthermore, the potential influence of this exclusion criterion on the generalizability of our identified optimal BP ranges should be considered. Relying on median BP values during hospitalization might not reflect hemodynamic variability, leading to potential misclassification due to BP fluctuations in sepsis. Future research should explore continuous BP modeling for a more accurate assessment of its impact on patient outcomes. Conclusions This single-center study suggests an optimal BP range for elderly sepsis patients that is associated with reduced 28-day mortality. These findings point to a potential optimal range for hemodynamic management in this population and warrant validation in prospective, multi-center studies. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Acknowledgements We would like to acknowledge all the study investigators for conducting this study. Abbreviations Blood Pressure Systolic Blood Pressure Diastolic Blood Pressure Mean Arterial Pressure Medical Information Mart for Intensive Care-IV Restricted Cubic Spline Propensity Score Matching Inverse Probability Weighting Doubly Robust Kaplan–Meier Hazard Ratio Confidence Interval Intensive Care Unit Atrial Fibrillation Temperature Heart Rate Respiratory Rate Oxygen Saturation White Blood Cell count Hemoglobin Platelet count Blood Urea Nitrogen Glasgow Coma Scale Sequential Organ Failure Assessment Mechanical Ventilation Acute Kidney Injury Kidney Disease: Improving Global Outcomes Multiple Imputation Interquartile Range Least Absolute Shrinkage and Selection Operator Gradient Boosted Model Standardized Mean Difference Surviving Sepsis Campaign Renal Replacement Therapy Institutional Review Board Author contributions YY: Data curation, writing – original draft; writing – review and editing. LC: Writing – original draft. LL: Data curation; supervision. FL: Data curation; supervision; picture and table editing. WL\YX: Data curation. BY: Writing – original draft; writing – review and editing. Funding No Funding. Data availability Data is provided within the manuscript or supplementary information files. Declarations Ethics approval and consent to participate This study utilized data from the Medical Information Mart for Intensive Care (MIMIC) database (version 2.0), which is a publicly available de-identified database containing retrospective healthcare data from critically ill patients. Individual patient consent was waived due to the retrospective and anonymized nature of the data. The first author completed the required training (Record ID: 43546933) and signed the Data Use Agreement to access MIMIC. This secondary analysis of de-identified data was deemed exempt from additional ethical review by the IRB of Wenzhou People’s Hospital. Throughout the study, we strictly adhered to the MIMIC data usage guidelines, ensuring no attempt to re-identify individuals or redistribute raw data. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Data Availability Statement Data is provided within the manuscript or supplementary information files.